Evidence about what works, for whom, and for how long is often lacking in international development. The impact of many interventions is simply unknown, while others continue to receive funding despite having been evaluated and found wanting. GiveDirectly was created because there is overwhelming evidence that cash transfers, in contrast, have a wide range of positive impacts.

Until recently, evidence of the long-term impacts of anti-poverty interventions has been particularly scant. Again, we have relatively good evidence for cash transfers; studies in Uganda and Sri Lanka find large impacts of lump-sum transfers on earnings 4-5 years later. But we think there is much more to be learned on two fronts.

First, we want to learn for whom one-time transfers have long-term effects. As several leading development economists (including two of our founders) discussed in a recent SSIR post, we would not expect one-time transfers to have long term impacts for all recipients. Yet we do not have enough data to predict for whom they do – whether farmers or factory workers, the young or the old, etc.

Second, we want to learn how long these effects last. To be clear, we would not expect them to last forever. A one-time grant can have a long-term impact if it lets the recipient make investments that she otherwise would not have been able to make. But even in places with weak financial systems, people without grants will likely find other ways to finance these same investments eventually, whether by saving their own money or borrowing someone else’s. Eventually we would expect their trajectories to converge, much as poorer but otherwise similar regions have tended to converge with richer ones over time. The question is how long convergence requires.

To answer questions like these, we need to be able to run big, long-term experiments. The math is simple: to learn about impacts on a group that’s only 1/5th of the population, for example, we need around 5 times the sample size we would otherwise need. We also need careful tracking, so that researchers can continue to find and survey study participants years down the road.

Some of the early studies at GiveDirectly may help a bit on these fronts, though they were largely optimized for other goals. The aspirations evaluation, for example, focused on the interplay between cash transfers and a video intervention and so allocated only 25% of villages to receive pure cash transfers. Our first study in Rarieda, Kenya focused on testing design variations (transfer sizes, timing, and recipients) and had a relatively small sample (the largest we could afford at the time); we expect to learn more when the results of a second endline are released, but less about subgroups or about the transfer design that we use today.

We’ve intentionally designed our most recent study to get around these constraints. It delivered transfers to a very large sample – over 650 villages – and every household enrolled received the lump sum ~$1000 transfers that are our current default. The independent research group (IPA) conducting the measurement have also followed best practices to ensure they can carefully track participants in the future.

We expect to get 18 month results back in the next few months. We’ll find out how $1,000 cash transfers change a range of outcomes, including economic, health, education, food security, wellbeing and female empowerment indicators. Crucially, the scale of this study means we’ll learn how different kinds of recipients use money and for whom the impacts are greatest. In the future, funders could potentially use evidence like this to precision-target cash transfers maximize the impacts they are most interested in, if they so wish.

Michael Cooke is GiveDirectly’s Research Director.

Back to List